AI - Conversational Agentic Commerce

The Problem

Our goal was to redefine how users discover and interact with ecommerce platforms by creating a more intuitive, conversational purchasing experience. We explored very early concepts in agentic commerce that reduced friction in product discovery, simplified purchasing workflows, and enabled intelligent agents to complete meaningful tasks on behalf of users with minimal interaction.

We envisioned systems capable of orchestrating complex, real-world actions autonomously — such as identifying close friends and family, coordinating baby shower invitations, and seamlessly sharing a personalized registry — creating experiences that felt less like traditional software and more like a proactive digital assistant.

Research & Discovery

Our research combined customer surveys, in-depth interviews, contextual inquiry, and exploratory concept validation. Because few existing systems aligned with the ambitious direction we were pursuing, we also had to act as visionaries — developing storyboards and future-state prototypes across multiple industries to gather meaningful public feedback and test emerging interaction models.

One of the most valuable aspects of our process was a “Build-to-Learn” methodology, where rapid iteration became a core driver of discovery. With each release, we spent countless hours analyzing customer interactions and refining language model responses. Observing not only what users asked, but how they naturally communicated with intelligent systems, provided remarkable insight into human behavior, intent, and conversational patterns.

Design & Iteration

We initially embraced fully open-ended conversational interaction, excited by the potential of natural language interfaces. However, through early implementation and user testing, we quickly recognized that the language models available at the time were not yet mature enough to reliably support unrestricted conversation at scale.

Two key challenges emerged immediately: users naturally communicated with complex sentence structures, and many requests required layered reasoning and multi-step contextual understanding beyond the capabilities of early models. In response, we rapidly evolved our strategy from fully open conversation to a more structured, guided conversational framework.

Rather than allowing unlimited conversational paths from the outset, we focused first on solving high-frequency customer needs — such as order status inquiries — and systematically expanded into increasingly sophisticated workflows over time. We designed guided conversational experiences that reduced ambiguity and improved reliability, while still preserving opportunities for users to move beyond predefined flows.

Maintaining this balance between structured interaction and conversational freedom became a critical part of our research strategy. While guided flows reduced friction and improved usability, open-ended interactions continued to generate invaluable insights into customer behavior, intent, and emerging expectations for AI-driven experiences.

Outcome & Impact

The project delivered an immediate reduction in calls to our Customer Service teams, significantly lowering operational support costs. By enabling customers to independently access information regarding their orders and account memberships, the volume of routine inquiries decreased substantially. This allowed Customer Service agents to focus on more complex customer issues, ultimately improving service quality and overall customer experience.